Documentation/content/world/entities/npc/npc-ai.en.md

title, type, weight

title	type	weight
NPC AI	docs	3

The NPC AI system provides intelligent decision-making through blackboards, decision makers, sensors, and instructions.

Packages:

• com.hypixel.hytale.server.npc.blackboard
• com.hypixel.hytale.server.npc.decisionmaker
• com.hypixel.hytale.server.npc.sensorinfo
• com.hypixel.hytale.server.npc.instructions

Blackboard System

The Blackboard is a shared memory space where NPC components communicate through key-value pairs.

Blackboard Class

public class Blackboard {
    private Map<BlackboardKey<?>, Object> data;

    // Store value
    public <T> void set(BlackboardKey<T> key, T value);

    // Retrieve value
    public <T> T get(BlackboardKey<T> key);
    public <T> T getOrDefault(BlackboardKey<T> key, T defaultValue);

    // Check existence
    public boolean has(BlackboardKey<?> key);

    // Remove value
    public void remove(BlackboardKey<?> key);

    // Clear all
    public void clear();
}

BlackboardKey

Type-safe keys for blackboard access:

// Predefined keys
public class BlackboardKeys {
    public static final BlackboardKey<Entity> TARGET =
        new BlackboardKey<>("target", Entity.class);
    public static final BlackboardKey<Vector3d> HOME_POSITION =
        new BlackboardKey<>("home_position", Vector3d.class);
    public static final BlackboardKey<Float> ALERT_LEVEL =
        new BlackboardKey<>("alert_level", Float.class);
    public static final BlackboardKey<Boolean> IN_COMBAT =
        new BlackboardKey<>("in_combat", Boolean.class);
}

// Custom keys
BlackboardKey<String> CUSTOM_KEY = new BlackboardKey<>("custom_data", String.class);

Using the Blackboard

NPCEntity npc = // get NPC
Blackboard bb = npc.getBlackboard();

// Set target
bb.set(BlackboardKeys.TARGET, targetEntity);

// Get home position
Vector3d home = bb.getOrDefault(BlackboardKeys.HOME_POSITION, npc.getPosition());

// Check combat status
if (bb.getOrDefault(BlackboardKeys.IN_COMBAT, false)) {
    // Handle combat
}

Decision Maker System

The Decision Maker evaluates options and selects the best action for the NPC to take.

DecisionMaker Interface

public interface DecisionMaker {
    // Evaluate and select best option
    Option evaluate(NPCEntity npc, Blackboard blackboard);

    // Get all available options
    List<Option> getOptions();

    // Add option
    void addOption(Option option);
}

Option Class

Options represent possible actions:

public class Option {
    private String id;
    private Evaluator evaluator;
    private Action action;
    private float basePriority;

    // Calculate score based on context
    public float evaluate(NPCEntity npc, Blackboard blackboard);

    // Execute the action
    public void execute(NPCEntity npc, Blackboard blackboard);
}

Evaluator Interface

Evaluators calculate option scores:

public interface Evaluator {
    // Return score from 0.0 to 1.0
    float evaluate(NPCEntity npc, Blackboard blackboard);
}

// Built-in evaluators
public class Evaluators {
    // Returns 1.0 if target exists
    public static final Evaluator HAS_TARGET = (npc, bb) ->
        bb.has(BlackboardKeys.TARGET) ? 1.0f : 0.0f;

    // Returns health percentage
    public static final Evaluator HEALTH_PERCENT = (npc, bb) ->
        npc.getHealth() / npc.getMaxHealth();

    // Returns 1.0 if at home
    public static final Evaluator AT_HOME = (npc, bb) -> {
        Vector3d home = bb.get(BlackboardKeys.HOME_POSITION);
        return npc.getPosition().distance(home) < 5.0 ? 1.0f : 0.0f;
    };
}

Creating a Decision Maker

DecisionMaker dm = new StandardDecisionMaker();

// Add attack option
dm.addOption(new Option(
    "attack",
    Evaluators.HAS_TARGET,
    new AttackAction(),
    10.0f  // High priority
));

// Add flee option
dm.addOption(new Option(
    "flee",
    (npc, bb) -> npc.getHealth() < 20 ? 1.0f : 0.0f,
    new FleeAction(),
    15.0f  // Higher priority when triggered
));

// Add wander option
dm.addOption(new Option(
    "wander",
    (npc, bb) -> 0.3f,  // Low constant score
    new WanderAction(),
    1.0f   // Low priority fallback
));

npc.setDecisionMaker(dm);

Sensor System

Sensors gather information about the world and update the blackboard.

Sensor Interface

public interface Sensor {
    // Process sensor input
    void sense(NPCEntity npc, Blackboard blackboard, float deltaTime);

    // Get sensor type
    String getSensorType();
}

SensorInfo Classes

Sensor information containers:

// Visual detection info
public class VisualSensorInfo {
    private List<Entity> visibleEntities;
    private float detectionRange;
    private float fieldOfView;

    public List<Entity> getVisibleEntities();
    public boolean canSee(Entity entity);
}

// Audio detection info
public class AudioSensorInfo {
    private List<SoundEvent> heardSounds;
    private float hearingRange;

    public List<SoundEvent> getHeardSounds();
    public Vector3d getLoudestSoundPosition();
}

// Threat detection info
public class ThreatSensorInfo {
    private List<Entity> threats;
    private Entity primaryThreat;
    private float threatLevel;
}

Built-in Sensors

// Visual sensor - detects visible entities
public class VisualSensor implements Sensor {
    private float range = 20.0f;
    private float fov = 120.0f;  // degrees

    @Override
    public void sense(NPCEntity npc, Blackboard bb, float dt) {
        List<Entity> visible = findVisibleEntities(npc);
        bb.set(BlackboardKeys.VISIBLE_ENTITIES, visible);
    }
}

// Proximity sensor - detects nearby entities
public class ProximitySensor implements Sensor {
    private float range = 5.0f;

    @Override
    public void sense(NPCEntity npc, Blackboard bb, float dt) {
        List<Entity> nearby = findNearbyEntities(npc, range);
        bb.set(BlackboardKeys.NEARBY_ENTITIES, nearby);
    }
}

// Damage sensor - reacts to damage taken
public class DamageSensor implements Sensor {
    @Override
    public void sense(NPCEntity npc, Blackboard bb, float dt) {
        if (npc.wasRecentlyDamaged()) {
            bb.set(BlackboardKeys.LAST_ATTACKER, npc.getLastAttacker());
            bb.set(BlackboardKeys.ALERT_LEVEL, 1.0f);
        }
    }
}

Instruction System

Instructions define high-level behavioral goals.

Instruction Interface

public interface Instruction {
    // Check if instruction should activate
    boolean shouldActivate(NPCEntity npc, Blackboard blackboard);

    // Execute instruction logic
    void execute(NPCEntity npc, Blackboard blackboard, float deltaTime);

    // Check if instruction is complete
    boolean isComplete(NPCEntity npc, Blackboard blackboard);

    // Get priority
    float getPriority();
}

Built-in Instructions

// Wander instruction
public class WanderInstruction implements Instruction {
    private float wanderRadius;
    private float minWaitTime;
    private float maxWaitTime;

    @Override
    public boolean shouldActivate(NPCEntity npc, Blackboard bb) {
        return !bb.has(BlackboardKeys.TARGET);
    }
}

// Guard instruction
public class GuardInstruction implements Instruction {
    private Vector3d guardPosition;
    private float guardRadius;

    @Override
    public void execute(NPCEntity npc, Blackboard bb, float dt) {
        if (npc.getPosition().distance(guardPosition) > guardRadius) {
            bb.set(BlackboardKeys.MOVE_TARGET, guardPosition);
        }
    }
}

// Follow instruction
public class FollowInstruction implements Instruction {
    private Entity followTarget;
    private float followDistance;
}

// Patrol instruction
public class PatrolInstruction implements Instruction {
    private List<Vector3d> patrolPoints;
    private int currentPoint;
}

Combining AI Systems

public void setupNPCAI(NPCEntity npc) {
    // Configure blackboard
    Blackboard bb = npc.getBlackboard();
    bb.set(BlackboardKeys.HOME_POSITION, npc.getPosition());
    bb.set(BlackboardKeys.AGGRO_RANGE, 15.0f);

    // Add sensors
    npc.addSensor(new VisualSensor(20.0f, 120.0f));
    npc.addSensor(new ProximitySensor(5.0f));
    npc.addSensor(new DamageSensor());

    // Configure decision maker
    DecisionMaker dm = new StandardDecisionMaker();
    dm.addOption(new Option("attack", hasHostileTarget, attackAction, 10.0f));
    dm.addOption(new Option("flee", lowHealth, fleeAction, 15.0f));
    dm.addOption(new Option("patrol", isGuard, patrolAction, 5.0f));
    dm.addOption(new Option("idle", always, idleAction, 1.0f));
    npc.setDecisionMaker(dm);

    // Add instructions
    npc.addInstruction(new GuardInstruction(guardPost, 10.0f));
    npc.addInstruction(new ReactToThreatInstruction());
}

Best Practices

{{< callout type="info" >}} AI Guidelines:

• Use the Blackboard for all inter-component communication
• Keep Evaluators simple and fast - they run frequently
• Use appropriate sensor ranges to balance awareness vs performance
• Design Instructions to be interruptible
• Test AI behavior with various scenarios {{< /callout >}}

{{< callout type="warning" >}} Performance: Large numbers of NPCs with complex AI can impact performance. Consider:

• Reducing sensor update frequency for distant NPCs
• Using LOD (Level of Detail) for AI complexity
• Limiting pathfinding requests {{< /callout >}}